HANCOTHON at MICCAI¶

Abstract¶

Clinical endpoints, such as predicting survival and risk of recurrence, are crucial in guiding oncological treatment decisions. In practice, clinicians combine diverse patient information to tailor treatment strategies that are most effective for each individual. However, classical AI tools often fall short because they rely on a single source of data, which limits their utility in complex clinical scenarios, particularly in treatment planning and decision-making. The novel HANCOCK dataset addresses this gap by providing a comprehensive multimodal resource, encompassing six data types: clinical data, pathology reports, histopathology images (of primary tumors and lymph nodes), tissue microarrays, tabular blood data, and free-text surgery reports. With data from 763 head and neck cancer patients collected at a single center to minimize technical biases, HANCOCK places the focus squarely on patient-specific insights. This challenge invites participants to harness the full potential of these diverse modalities, thereby maximizing their predictive power for critical endpoints, such as survival and recurrence prediction, in precision oncology. Challenge participants must devise strategies for integrating the various data streams and combining large image datasets with structured and unstructured text data to perform two binary classification tasks. The two tasks use the same data to (i) determine the 5-year-survival to potentially help in therapy and adjuvant therapy prediction and (ii) estimate the recurrence risk to ensure timely follow-up intervals for monitoring purposes.

Task¶

Classify the 5-year survival and 2-year recurrence for head and neck cancer patients. The evaluation will be done based on accuracy and F1-score. Each task will be evaluated separately.

Who can Participate¶

You can participate alone or gather your friends and colleagues to devise a team solution.

Dataset¶

We provide a comprehensive multimodal dataset that includes tabular data, histopathology images, and pre-extracted embeddings from histopathology images. As a participant, you can freely decide which of these data types you want to use for your best solution. The following data types will be available in the challenge:

• Structured Blood Data
• Structured Pathology Data
• Structured Clinical Data without the target-related columns
• Text Data in English
• Pre-extracted tumor center cores from tissue microarrays
• Pre-extracted embeddings of the primary tumor whole slide images generated with UNI
• Pre-extracted embeddings of the lymph node whole slide images generated with UNI

How to Participate¶

The challenge itself is hosted on Grand Challenge, where your final solution will be tested. The training data for your model can be downloaded at our HANCOCK research website. Before downloading, please verify that you are using only the modalities listed in the Dataset chapter. The winners will be announced at MICCAI 2025 (23rd - 27th September 2025).
If you have problems getting started, consider joining our Slack workspace or meeting us in one of your upcoming Q&A Sessions.

Prices¶

Prices will only be granted to the Closed Development Phases. The Open Development Phases are only for testing your algorithm within Grand Challenge.

2-year recurrence:

• 🥇 1st place: 200€
• 🥈2nd place: 150€
• 🥉3rd place: 50€

5-year survival:

• 🥇 1st place: 200€
• 🥈2nd place: 150€
• 🥉3rd place: 50€

Members from the same department as the organizer may participate, but are not eligible for awards.

Communication¶

If you want our support or help with this challenge, you can attend one of the upcoming Q&A Sessions and join our Slack channel here.

Publication Policy¶

Given the interest in multimodal patient outcomes and risk prediction, we plan to coordinate a publication that summarizes the main approaches and insights derived from the challenge. This will be submitted to renowned, peer-reviewed journals, such as Medical Image Analysis (MedIA) or journals within the Nature family. We will offer co-authorship to participants.

Additionally, we impose no embargo period, allowing participants to publish their work freely.